Social networks are a set of individuals or groups and interactions between them. These interactions could be friendship activities or trade relations. Social networks play an important role in information diffusion and maximizing the influence on users. In?uence maximization is the problem of ?nding a small subset of nodes (seed nodes) in a social network that could maximize the spread of in?uence. Therefore the goal is to choose the initial set of users to maximize the diffusion process. Information diffusion in social networks occurs under a diffusion model on a graph whose edges are labeled with probabilities of in?uence between nodes. Real world social networks are unweighted and the probabilities of influence between nodes (edge weights) are unknown. The main idea of this thesis is estimating the influence probability from real data in order to maximize information diffusion in network. The influence between each pair of users is obtained by measuring the transfer entropy or information transfer as defined in information theory. This method measures causal rate between time series of nodes activity time in network. The results show that our estimated diffusion rate of the whole network and influential nodes are different from the other methods. In a real social network, messages are sent or received with a lag. Here, we used the transfer entropy measure to estimate the delay in the transmission of information. Our goal is to measure the delayed transfer entropy between the interactions of network nodes on each other. The validity of the proposal method is assessed on real world and synthesized data sets, shows improvements in the capability of estimating influence probabilites and ranking the in?uential nodes over existing methods. Keywords: Social network, information diffusion, influence maximization, causality
